Bug 756434

Created attachment 535517 [details]
config file for network

You don't give the version/origin of the libvirt you're using, but I'm assuming you've built the latest sources from git.

You also don't give the platform or release of that platform, but based on the machine type in the domain XML I assume it is RHEL 6.1.

If those are correct, then the problem is that macvtap passthrough mode is not supported until at least kernel-2.6.32-175.el6 / RHEL6.2. Other macvtap modes should be supported, though - have you tried bridge or private?

If you really need macvtap passthrough mode, you should try to get hold of a RHEL6.2 beta.

Also, for a problem like this encountered while using a build from upstream, you can probably get quicker results by first sending mail to libvir-list. The developers pay much closer attention to this than to the upstream bug tracker (since most items in the upstream bug tracker are first brought up on the mailing list).

Please let me know if my assumptions about the origin of your libvirt and the release of OS you're using are correct. I would also like to hear if using a different macvtap mode works correctly.

(In reply to comment #2)

> Please let me know if my assumptions about the origin of your libvirt and the
> release of OS you're using are correct. I would also like to hear if using a
> different macvtap mode works correctly.

Thankyou for your quick response
Your assumptions about the origin of my libvirt and about my OS are correct.
I will update my kernel version.

I tried using the mode vepa and I see the following error:
[root@c6100r sshah]# virsh start dibenchvm
error: Failed to start domain dibenchvm
error: Cannot set interface flags on 'macvtap0': Device or resource busy


I use /var/log/libvirt/libvirtd.log to get all the debug output (log_level is set to 1 in /etc/libvirt/libvirtd.conf)

libvirtd.log shows the following output:

2011-11-24 10:46:33.932: 2139: debug : networkAllocateActualDevice:2963 : Using physical device eth2, usageCount 1
2011-11-24 10:46:33.932: 2139: debug : openMacvtapTap:288 : openMacvtapTap: VM OPERATION: create
2011-11-24 10:46:33.934: 2139: debug : vpAssociatePortProfileId:1104 : Associating port profile '(nil)' on link device 'macvtap0'
2011-11-24 10:46:33.934: 2139: debug : vpAssociatePortProfileId:1106 : vpAssociatePortProfileId: VM OPERATION: create
2011-11-24 10:46:33.934: 2139: error : virNetDevSetOnline:454 : Cannot set interface flags on 'macvtap0': Device or resource busy
2011-11-24 10:46:33.934: 2139: debug : vpDisassociatePortProfileId:1163 : Disassociating port profile id '(nil)' on link device 'macvtap0' 
2011-11-24 10:46:33.934: 2139: debug : vpDisassociatePortProfileId:1165 : vpDisassociatePortProfileId: VM OPERATION: create
2011-11-24 10:46:33.959: 2138: debug : virEventPollRunOnce:629 : Poll got 1 event(s)
2011-11-24 10:46:33.959: 2138: debug : virEventPollDispatchTimeouts:414 : Dispatch 5
2011-11-24 10:46:33.959: 2138: debug : virEventPollDispatchHandles:459 : Dispatch 8
2011-11-24 10:46:33.959: 2138: debug : virEventPollDispatchHandles:473 : i=0 w=1
2011-11-24 10:46:33.959: 2138: debug : virEventPollDispatchHandles:473 : i=1 w=2
2011-11-24 10:46:33.959: 2138: debug : virEventPollDispatchHandles:473 : i=2 w=3
2011-11-24 10:46:33.959: 2138: debug : virEventPollDispatchHandles:473 : i=3 w=4
2011-11-24 10:46:33.959: 2138: debug : virEventPollDispatchHandles:473 : i=4 w=5
2011-11-24 10:46:33.959: 2138: debug : virEventPollDispatchHandles:473 : i=5 w=6
2011-11-24 10:46:33.959: 2138: debug : virEventPollDispatchHandles:487 : EVENT_POLL_DISPATCH_HANDLE: watch=6 events=1
2011-11-24 10:46:33.959: 2138: debug : udevEventHandleCallback:1467 : udev action: 'add'
2011-11-24 10:46:33.959: 2138: debug : udevGetDeviceProperty:120 : udev reports device 'macvtap0' does not have property 'DRIVER'
2011-11-24 10:46:33.959: 2138: debug : udevGetDeviceType:1095 : Found device type '(null)' for device 'macvtap0'
2011-11-24 10:46:33.959: 2138: debug : udevGetDeviceProperty:120 : udev reports device 'macvtap0' does not have property 'PCI_CLASS'
2011-11-24 10:46:33.959: 2138: debug : udevGetDeviceProperty:140 : Found property key 'INTERFACE' value 'macvtap0' for device with sysname 'macvtap0'
2011-11-24 10:46:33.959: 2138: debug : udevGetDeviceProperty:140 : Found property key 'INTERFACE' value 'macvtap0' for device with sysname 'macvtap0'
2011-11-24 10:46:33.959: 2138: debug : udevGetDeviceSysfsAttr:209 : udev reports device 'macvtap0' does not have sysfs attr 'address'
2011-11-24 10:46:33.959: 2138: debug : udevGetDeviceSysfsAttr:209 : udev reports device 'macvtap0' does not have sysfs attr 'addr_len'

The steps to reproduce are the same as in the Description.

Shradha,

I'm assuming from the patch you posted to libvir-list that updating your kernel solved the problem with passthrough mode. Is that correct?

As far as your problems with vepa, I'm unable to test vepa because it requires a vepa-capable switch. Your best bet for getting help with that would be to post a message to libvir-list, where the developers who did that work will see the problem and may be able to provide some help. Very few people see problem reports in the upstream bugzilla for libvirt, so it's not likely to get much response here, Upstream libvirt's bugzilla tends to mostly get used as a place to hold feature requests, while bug reports of this type are usually filed against a specific Linux distro, or handled completely on libvir-list.

(In reply to comment #5)
> Shradha,
> 
> I'm assuming from the patch you posted to libvir-list that updating your kernel
> solved the problem with passthrough mode. Is that correct?

Laine,

Yes, updating the kernel solved my problem with the passthrough mode.

I will post my message regarding vepa mode to libvir-list.

Many Thanks for all your help.

Okay, I'm closing this bug then, since the originally reported problem is a kernel problem that's already fixed (both upstream and in the next RHEL release).